overclocking my 3500x with Asrock B550M steel legend.is that Voltage behaviour right?

N

nolive72

Member
First of all, I want to say its not a rant against this great piece of software that HWINFO is, obviously;)

I have been OCing since the days of the Intel G3258 and always used fixed voltage to proceed, being on a 4670k afterwards, a Ryzen 1500X 3 yrs ago and now the 3500X I am using on my main Gaming rig

For some reason this weekend I decided to test a Voltage offset feature in my MB, the ASROCK B550M Steel legend. So far I had 3 OC/UV preset I had tested in and out as stable in benchmark and heavy gaming. all these OC were set in Ryzen Master, not in the BIOS

  1. All core OC 4.1Ghz @1.13V
  2. All core OC 4.4Ghz @1.2375V
  3. All core OC 4.5Ghz @1.35V
I am running my rig with a 240mm AIO and all temps were nicely in check, even the extreme OC @1.35V the CPU wont go higher than 75degC in intensive benchmark like AIDA64 or CB R23

the 1st run in the pic below is just leaving Voltage in AUTO in the BIOS and run CB R23 to get temps and Voltage readings monitoring on focus

CPU boosting to 4,1Ghz so normal, CPU Voltage SVI2, CPU Core VID effective and Vcore all giving me same reading around 1.36/1.38V with a Max CPU power package around 72W and temps in low 70s, again nothing to worry about in my views, well within the chip limits

1st run Stock in Bios and in Ryzen master no touch .jpg


the 2nd run so I did to test the Offset Voltage feature was a -100mV in the OCtweaker section in the BIOS and boot into Windows

same testing with CB R23 and to my surprise, CPU Voltage SVI2, CPU Core VID effective were not affceted by the offset voltage, only the Vcore saw a decrease

all temps, power consumption were actually positively affected by the change too

2nd run -100mV in Bios and in Ryzen master no touch .jpg


then final run, while in Windows, I applied an OC which is to boost all core 4.1Ghz @1.23V demand in Ryzen Master

in that case, CPU Voltage SVI2, CPU Core VID effective were reflecting my demand Voltage but the Vcore dropped to 1.13V basically applying, at least that's my understanding, the Ryzen Master voltage minus the 100mV that I set in the BIOS as Offset voltage

2nd run -100mV in Bios and in Ryzen master 4.1Ghz 1.23V .jpg


I have always read that the CPU Voltage SVI2, CPU Core VID effective were the ones Voltage to believe the CPU was actually getting so I am little bit confused now and looking for someone who could bring some clarity here

Maybe that's an ASROCK BIOS thing or the normal behaviour of these 3000series Ryzen or maybe me doing the wrong assessment, so again not a rant at all against HWINFO readings, its just me trying to understand a phenomena to be in better control of my Rig

thanks so much
 
I don't know what else to add, you have already been told that the most reliable value is the "CPU Core Voltage (SVI2 TFN)". Vcore is affected by other effects (depends on sampling point) and often not reliable.
 
thank you Martin

I was just referring to translate Voltage to power, and the generated heat obviously, and it seems to me that my rig behaves more like Power=Vcore*Current when I look into HWINFO output of the 3 cases I have tested

Also I would like to better understand what an offset Voltage in my ASROCK BIOS don't translate to SVI2 and Core VID effective voltages

I will post in more MB related forum to see if people have similar findings or if I am doing something wrong with my input in the BIOS

thanks again for your reply in any case and of course for the the software you share to the community
 
thanks both for your replies. I am posting on REDDIt ASROCK and OCing subs to see if I can get people who experience the same discrepancy between Voltage offset in Bios not reflecting in Windows.

I look at this Power deviation reporting numbers and see HWINFO giving me 80% so is my understanding correct that its underestimating the power actually consumed by the chip voltage I am applying? and that's potentially a bad thing?
 
nolive72 said:
I look at this Power deviation reporting numbers and see HWINFO giving me 80% so is my understanding correct that its underestimating the power actually consumed by the chip voltage I am applying? and that's potentially a bad thing?
Click to expand...
No it doesn't mean that it can hurt the CPU if that is what you mean. Just that it makes it operate beyond stock limits.
If and only if everything is on stock settings (CPU voltage and multi) then the CPU still has its internal protection (silicon FITness management) that would not allow any damage.
When you take things at manual (CPU voltage and multi) this FIT is disabled. PBO settings don't disable the FIT.
 
Last edited:
not quite, what I meant is that as Power=Voltage*Current I am concerned that underestimated power due to ASROCK strategy with this board Power reporting deviation could mean that either Voltage or Current, or both, are actually underestimated in HWINFO, and the CPU would get more than what's safe hence reducing its lifespan

These days, what I am running is what I called my OC Nr1
  1. All core OC 4.1Ghz @1.13V
to be on the safe side and the Chip delivers enough performance for my needs so I could leave this as it is until I have some more quality time to revisit the topic with more testing

thanks again for your input, it was helpful
 
We are saying the same thing.

When the PRD reports 80% (under full load) it means that that the reported total power consumption (PPT) of the CPU package is only the 80% of the true power consumption.

Random example:
If you see a PPT of 70W and PRD 80% it means that the true power is

70 / 0.8 = 87.5

The true power consumption is 87.5 W instead of the reported 70W. That happens because the CPU thinks it draws less than true, and it raises voltage and speed to match its stock limit.
So if your CPU has a stock limit of 70W this makes it work at 87.5W
Even though it is exceeding stock limits it can’t hurt it self because the silicon FITness management is still in play. It is just a stock power consumption limit override and not an exceed in silicon integrity/longevity limit.
It’s different.
 
yes absolutely making sense

and that's why I tested undervolting my chip because I was not happy to see the board throwing above 1.3V to run it at 4.1Ghz.

I did some more testing last night and it seems the Chip can now do 4.2Ghz @1.1375V in most of my Gaming scenarios without crashing so that could become my daily driver, instead of Stock+Power saving Windows Plan as I have it today

Now of course, I will set Manual fixed Voltage/clocks either in BIOS or Ryzen Master but I guess there is no way to get the Voltage below this 1.1375V when the chip is idling right?
I have tried the Windows and Ryzen power Plan(s) and if I see downclocks, the Voltage stays there

thanks so much
 
Always better to use BIOS settings.

On auto (stock) boost, ZEN2/3 when idles or having very light loads (browsing, watching vids) it can be supplied up to 1.45~1.5V. When threads and load goes up (current (EDC) goes up) the CPU is dropping voltage to 1.1~1.3V depending on the model and the type of load. This is absolutely typical and normal behavior. Because 3500X is a 6core/6threaded CPU cant be loaded too much, like others 6core/12threaded models (3600/3600X). So seeing 1.3V under full load is ok.

But if you want to run it @4.2GHz/1.1375V its still ok as long as EDC/PPT is not exceeding by too much the stock limits of 3500X under your worst case scenario load.
There is a way to have different voltage depending load/idle through CPU Vcore LLC settings (Vdroop) but there is no point to decrease it more when idle. Usually (99% of static OC) users set a voltage for idle and make it drop when going full load (Vdroop under load). This is the normal and right behavior that you should seek.
 
thanks again.

yes I guess I will have to play with LLC, its the 1st MB I own that has such feature so I will need to read first and experiment then. again, much appreciated the know-how sharing:cool:
 
You must log in or register to reply here.

Similar threads

C
Intel 11900kf VID voltage values BIG spikes
Replies
4
Views
92
ctrl-art-del
C
N
CPU TLB Errors
Replies
4
Views
506
Nerozord
N
wwvierg
I too can't find loadline info (sorry) ... Thinkpad P1 Gen 6
Replies
2
Views
115
wwvierg
wwvierg
L
CPU High Voltage Spike?
Replies
2
Views
519
LouZ23
L
O
High Voltage and Clockspeeds when using HWinfo64
Replies
2
Views
89
O.F.I.
O
Back
Top